Statin Use and Risk for Primary Liver Cancer in the Clinical Practice Research Datalink

Abstract

Background:

Statins (3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors) are widely prescribed to reduce cholesterol levels. Studies have suggested that statins are associated with reduced risk for liver cancer, but much of the evidence is from regions of the world with high liver cancer incidence rates. The current study examined the statins–liver cancer relationship in a low-rate region and examined the effects of preexisting liver disease and diabetes on that association.

Methods:

A nested case-control study was conducted within the United Kingdom’s Clinical Practice Research Datalink (CPRD). Persons diagnosed with primary liver cancer between 1988 and 2011 were matched to controls at a four-to-one ratio. Matches stratified on liver disease and on diabetes were also completed. Odds ratios (ORs) and 95% confidence intervals (CIs) for associations of statins with liver cancer were estimated using conditional logistic regression.

Results:

In total, 1195 persons with primary liver cancer were matched to 4640 control patients. Statin use was associated with a statistically significantly reduced risk for liver cancer (OR _adj = 0.55, 95% CI = 0.45 to 0.69), especially among current users (OR _adj = 0.53, 95% CI = 0.42 to 0.66). The reduced risk was statistically significant in the presence (OR _adj = 0.32, 95% CI = 0.17 to 0.57) and absence of liver disease (OR _adj = 0.65, 95% CI = 0.52 to 0.81) and in the presence (OR _adj = 0.30, 95% CI = 0.21 to 0.42) and absence of diabetes (OR _adj = 0.66, 95% CI = 0.51 to 0.85).

Conclusions:

In the current study in a low-rate area, statin use was associated with a statistically significantly reduced risk for liver cancer overall. Risk was particularly reduced among persons with liver disease and persons with diabetes, suggesting that statin use may be especially beneficial in persons at elevated risk for liver cancer.

Primary liver cancer is the sixth most commonly occurring cancer in the world and because of a very poor prognosis, the second most frequent cause of cancer mortality ( 1 ). In the majority of high-rate liver cancer areas, mainly in Asia and Africa, the most common risk factors are chronic hepatitis B virus (HBV) infection and aflatoxin contamination of foodstuffs. In contrast, in low-rate areas, such as Europe and North America, the most common risk factors are excessive alcohol consumption, diabetes/obesity, hepatitis C virus (HCV) infection, and nonalcoholic fatty liver disease (NAFLD) ( 2 ). Incidence rates have been increasing in many low-rate regions ( 3 ), likely because of the increased prevalence of diabetes, obesity, NAFLD and HCV infection ( 4 ). Predictions of further increases in incidence ( 5 ) underscore the need to identify effective prevention strategies.

Statins (3-hydroxy-3-methylglutaryl coenzyme A (HMG-Co-A) reductase inhibitors) are commonly used cholesterol-lowering medications that have demonstrated effectiveness in the primary and secondary prevention of cardiovascular disease ( 6 ). Although statins were initially suspected of increasing the risk of cancer ( 7 ), subsequent examination failed to support those concerns ( 8 , 9 ) and raised the possibility that statins could have anticarcinogenic effects ( 10 ) related to inhibited angiogenesis, enhanced apoptosis, and metastasis inhibition ( 11 ). A potential for liver cancer prevention is particularly indicated, as the liver, the target organ for statins, sequesters the majority of the drug. Promising evidence that statins may decrease risk of liver cancer has been reported in observational studies, many of which were conducted in Taiwan ( 12–16 ). The results of studies from areas with low rates of liver cancer, however, have been less consistent ( 17–22 ). Furthermore, there has been debate about whether previously reported statins–liver cancer associations are because of biased prescribing patterns ( 23 ). Although rare, statin-related hepatotoxicity is not unknown ( 24 ), thus there may be a reluctance to prescribe statins to persons with preexisting liver disease. The extent to which prescribing bias has influenced the reported inverse association of statins and liver cancer is unclear. Stratification on liver disease in several studies ( 13 , 17 ) has provided some information on the topic, but more data are needed. More data are also needed on the effect of statins among persons with the most common risk factors, such as diabetes, in low-rate areas. Thus the current study sought to examine, in a low-rate area, the statins–liver cancer relationship overall and among persons with liver disease and diabetes.

Methods

A nested case-control study was conducted within the Clinical Practice Research Datalink (CPRD) of the United Kingdom (UK). The CPRD is a large, population-based, automated medical records database that contains information on approximately 8.5% of the UK population. The UK National Health Service (NHS) provides universal coverage, therefore no segment of the population is excluded from the CPRD and the age and gender distributions are representative of the general UK population ( 25 ). General practitioners (GPs) who contribute to the CPRD provide the data in an anonymous format for research purposes. All GPs have been trained to record demographic data, medical information, details of hospital stays, and deaths. Diagnoses, physical findings, symptoms, and administrative events, such as referrals to specialists, are recorded using Read codes rather than International Classification of Diseases (ICD) codes. Detailed information is available for all medications prescribed. Several studies have examined the validity of the information recorded in the CPRD and indicate that the data are reasonably complete and accurate with regard to clinical illnesses diagnosed either by the GP or a specialist ( 26 , 27 ). Specifically, it has been demonstrated that more than 90% of information from manual medical records gets recorded electronically ( 26 , 27 ) and approximately 95% of all electronically identified primary cancers are confirmed as incident cancers ( 28 ). The base population for the current study included all persons between the ages of 10 and 90 years in the CPRD between the years 1988 and 2011. The study was approved by the National Institutes of Health Human Research Protection Program.

Case Patients and Control Patients

The eligibility criteria for case patients were: 1) first-time diagnosis of primary liver cancer (Read codes B150300, B150z00, B152.00), 2) no prior diagnosis of the cancers most likely to metastasize to the liver: lung, stomach, breast, colon, or pancreatic cancer, and 3) no diagnosis of any other cancer (except nonmelanoma skin cancer) in the three years prior to the index date. The index date was defined as the date of liver cancer diagnosis minus one year. All case patients were required to have at least two years of recorded activity in the CPRD prior to the index date. Persons with any code for liver metastases were excluded from the study. Of the 1195 case patients included in the study, the majority (n = 1036, 86.7%) had supporting clinical codes that indicated presence of liver cancer such as diagnostic exams (biopsies), treatment (chemotherapy, radiotherapy, surgery), palliative care, and referrals to specialty care. The minority (n = 159, 13.3%) who had no supporting clinical codes were often persons who died shortly after the liver cancer diagnosis, prior to treatment, or persons whose cancer diagnosis was recorded at the time of death.

Control patients were matched to case patients at a four-to-one ratio on age (same year of birth as case), sex, general practice, index date (one year prior to case’s diagnosis date), and number of years in the CPRD prior to the case’s index date. Control patients had to be free of any cancer (except nonmelanoma skin cancer) prior to the index date of the matched case patient and were required to have at least two years of history in the CPRD prior to the case’s index date. Only three eligible control patients could be identified for 140 of the case patients, thus 1055 case patients have four control patients and 130 case patients have only three control patients.

Three separate case-control matches were completed for the study. For the full analysis, control patients were matched to all 1195 liver cancer case patients. For the chronic liver disease-matched analysis, control patients with liver disease were matched to the 170 case patients with liver disease, and control patients without liver disease were matched to the 1025 case patients without liver disease. For the diabetes-matched analysis, control patients with diabetes were matched to the 346 case patients with diabetes, and control patients without diabetes were matched to the 849 case patients without diabetes. The inclusion and exclusion criteria were the same across all three matches.

Exposure to Statins

For all analyses, statin use was defined as having two or more statin prescriptions recorded prior to the index date of the case patients and control patients. Nonuse was defined as one or no statin prescriptions prior to the index date. Current statin use was defined as use that ended within one year prior to the index date, while past use was defined as use that ended more than one year prior to the index date. Exposure to statins was examined both in relation to number of prescriptions dispensed and in relation to cumulative dose, which was defined as number of pills times the dose per pill.

In addition to analyzing statin exposure as a single entity, specific statins (atorvastatin, simvastatin, pravastatin, fluvastatin, rosuvastatin, cerivastatin, and the combination of ezetimibe with simvastatin) were examined individually. Lovastatin, a common statin in other countries, is not marketed in the United Kingdom.

Statistical Analysis

Prior to initiating the statistical analysis, the distributions of all covariables were examined to ensure there were no outlying values. For variables with missing values, “unknown” categories were created for the analyses. Interactions between the major matching factors and covariables and statin use were also examined. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using conditional logistic regression to assess crude and adjusted risk estimates for the relationship of statin use to liver cancer. Wald Chi-square tests were calculated to determine whether trends across categories were statistically significant. Known liver cancer risk factors for which adjustment was made in the analysis included: body mass index (BMI), smoking, alcohol-related disorders, HBV and/or HCV infection, diabetes, and rare metabolic disorders (hemochromatosis, Wilsons Disease, alpha-1 antitrypsin deficiency, porphyrias). As some evidence has suggested that aspirin ( 29 ) and antidiabetic medications ( 30 ) might be related to liver cancer risk, adjustment was made for those exposures. In addition, adjustment was made for paracetamol use because of a statistically significantly increased risk noted in the univariate analysis.

In addition to the full match, the liver disease match, and the diabetes match, two sensitivity analyses were conducted. The first analysis was restricted to case patients with clinical codes for treatment of liver cancer (eg, surgery, chemotherapy, or palliative care) and their matched control patients. The second analysis used an index date of two years prior to the case’s date of diagnosis rather than one year. All statistical tests were two-sided. P values of less than .05 were considered statistically significant.

Results

Table 1 displays the characteristics of the 1195 liver cancer case patients and 4640 control patients included in the overall analysis and the univariate analysis of known risk factors. Case patients and control patients were matched on index year, age at index year, sex, and length of enrollment in CPRD prior to index date, thus there were no differences in these factors. Not unexpectedly, case patients were statistically significantly more likely than control patients to be obese (BMI ≥ 30), to be current or ex-smokers, have an alcohol-related condition, be infected with HBV or HCV, have chronic liver disease, use paracetamol, have a rare metabolic disorder, and to have type I or type II diabetes.

Table 2 displays the relationship of statin use to liver cancer in the overall analysis. Statin use was associated with a statistically significantly reduced risk (OR _adj = 0.55, 95% CI = 0.45 to 0.69) of liver cancer. There was a statistically significant dose-response relationship between risk and cumulative daily dose of statins ( P < .001), which was mirrored by the relationship with number of statin prescriptions ( P < .001). When the data were stratified by recency of use, the statistically significant dose-dependent relationship was restricted to current users (OR _adj = 0.53, 95% CI = 0.42 to 0.66). Analysis by type of statin found statistically significant inverse associations for the two most commonly prescribed statins, simvastatin (OR _adj = 0.57, 95% CI = 0.45 to 0.74) and atorvastatin (OR _adj = 0.53, 95% CI = 0.38 to 0.75), as well as for rosuvastatin (OR _adj = 0.42, 95% CI = 0.19 to 0.97). The odds ratios for the other statins were of similar magnitude as the ORs for atorvastatin, simvastatin, and rosuvastatin, but did not attain statistical significance.

Table 3 displays the relationship of statin use and liver cancer among persons with and without chronic liver disease. Among persons with chronic liver disease, statin use was statistically significantly associated with reduced risk (OR _adj = 0.32, 95% CI = 0.17 to 0.57) in a dose-response manner such that risk declined with increasing cumulative daily dose of statins ( P < .001). The reduced risk was more evident among current (OR = 0.30, 95% CI = 0.16 to 0.58) than past (OR = 0.38, 95% CI = 0.12 to 1.18) users, although the numbers of past users were small. Among the persons without chronic liver disease, statin use was also statistically significantly associated with reduced liver cancer risk (OR _adj = 0.65, 95% CI = 0.52 to 0.81), with risk declining as cumulative dose increased ( P < .001). The association was seen among current (OR = 0.62, 95% CI = 0.50 to 0.78) but not among past statin users (OR = 0.92, 95% CI = 0.57 to 1.47).

Table 4 displays the relationship of statins and liver cancer among persons with and without diabetes. Among persons with diabetes, statin use was statistically significantly associated with reduced risk (OR _adj = 0.30, 95% CI = 0.21 to 0.42), with decreasing risk associated with increasing cumulative dose ( P < .001). The association was seen both among current (OR _adj = 0.29, 95% CI = 0.21 to 0.41) and past users (OR = 0.45, 95% CI = 0.21 to 0.95). Statin use was also statistically significantly associated with reduced risk among persons without diabetes (OR = 0.66, 95% CI = 0.51 to 0.85), which was evident among current (OR = 0.60, 95%CI 0.46 to 0.79) but not among past users (OR = 1.23, 95% CI = 0.69 to 2.19).

The results of the sensitivity analysis restricted to case patients with supporting clinical codes (86.7%) and their control patients did not differ from that in the overall analysis (OR = 0.66, 95% CI = 0.53 to 0.80) (data not shown). Similarly, the sensitivity analysis that was based on an index date of two years prior to the case’s diagnosis date rather than one year resulted in findings very similar to the main analysis (OR = 0.59, 95% CI = 0.48 to 0.72) (data not shown). The interaction analysis of covariables with statin use identified no statistically significant interaction effects.

Discussion

In the current study conducted in a low-rate region, statin use was associated with a statistically significantly reduced risk of liver cancer. The relationship was statistically significant in particular among current statin users. The reduced risks between atorvastatin, simvastatin, and rosuvastatin and liver cancer were statistically significant, while the inverse associations of pravastatin, fluvastatin, and cerivastatin did not attain statistical significance. Among persons at elevated risk of liver cancer, persons with liver disease, and persons with diabetes, statins were associated with particularly statistically significant reduced risks.

The results of the current study are similar to those of prior studies from Taiwan, all of which have used the Taiwan National Health Insurance Research Database ( 12–16 ). Examining 1166 liver cancer case patients and 1166 control patients, Chiu et al. ( 12 ) reported a statistically significant inverse association (OR = 0.53, 95% CI = 0.45 to 0.83), which was subsequently replicated by Leung et al. (OR = 0.44, 95% CI = 0.28 to 0.72) ( 14 ). Tsan et al. also reported statistically significant inverse associations among persons infected with HBV (OR = 0.47, 95% CI = 0.36 to 0.61) ( 16 ) and persons infected with HCV (OR = 0.53, 95% CI = 0.49 to 0.58) ( 15 ), while Lai et al. ( 13 ) reported statistically significant inverse associations for simvastatin (OR = 0.69, 95% CI = 0.50 to 0.94), lovastatin (OR = 0.52, 95% CI = 0.36 to 0.76), and atorvastatin (OR = 0.70, 95% = 0.53 to 0.93). Inverse associations were also seen for fluvastatin, pravastatin, and rosuvastatin, but the results were not statistically significant.

Prior results from low-rate liver cancer areas have been less consistent. In a Danish prospective study, Friis et al. reported a null association (OR = 1.16, 95% CI = 0.46 to 2.90) ( 19 ). Similarly, an analysis of a large US electronic medical records database by Marelli et al. reported a null association, although liver cancer occurred among 0.37% of non-statin users, but only among 0.10% of statin users ( 21 ). Conversely, support for an inverse association was found in two US studies conducted among the US Department of Veterans Affairs’ patient population. Both El-Serag et al. ( 17 ), studying men with diabetes (OR = 0.74, 95% CI = 0.64 to 0.87), and Khurana et al. ( 20 ), studying men with HCV infection (OR = 0.52, 95% CI = 0.41 to 0.67) reported statistically significantly reduced risks of liver cancer. In addition, two studies conducted among members of health maintenance organizations in the United States ( 18 , 22 ) reported statistically significantly reduced risks of liver cancer. The explanation for the inconsistency in results is not clear, although the small number of liver cancers (n = 24) in the Marelli et al. study, may have limited the ability to detect an effect. Two meta-analyses of statins and liver cancer have also been reported ( 31 , 32 ). While both meta-analyses concluded that statins were inversely related to liver cancer risk, one ( 32 ) suggested that the protection might be limited to Asian populations where HBV infection was a major factor. The results of the current study, however, suggest that statins are of equal benefit in non-Asian populations where HBV infection is not as dominant a risk factor.

In addition to observational studies, secondary analyses of randomized controlled trials (RCTs) of statins and cardiovascular disease have attempted to examine the risk of cancer. Of the few trials that had information on liver cancer, none have found a statistically significant association. Sato et al. ( 33 ) reported an observed/expected ratio of 0.63 (95% CI = 0.01 to 3.49) from an RCT that included 179 participants in the statins arm, one of whom developed liver cancer. Matsushita et al. ( 34 ), in a meta-analysis of three Japanese RCTs, reported a hazard ratio of 0.58 (95% CI = 0.18 to 1.84) based on five liver cancers in the statins arm and seven in the control arm. Emberson et al. ( 35 ), in a meta-analysis of cancer outcomes in 27 RCTs, reported no relationship between statin use and liver cancer ( P = .39), based on 42 liver cancers in the statins arm and 51 in the control arm. Overall, the information from the secondary analysis of cardiovascular disease RCTs is limited by the fact that liver cancer is a rare outcome that takes years to develop. In addition, none of the trials adjusted for major liver cancer risk factors.

Whether certain statins have different effects on risk of liver cancer is not certain, although the hydrophilic statins such as pravastatin and rosuvastatin are more hepatoselective than are the lipophilic statins such as simvastatin, atorvastatin, lovastatin, and fluvastatin ( 11 ). Nevertheless, the current study found that members of both groups were associated with reduced risk. In addition to the current study, three other studies have examined the effect of individual statins ( 12 , 13 , 16 ) and all four have found statistically significantly reduced risk with use of atorvastatin or simvastatin. In addition, in the study of El-Serag et al. ( 17 ), simvastatin was the only statin able to be examined independently and was related to statistically significantly reduced risk. Rosuvastatin has also had a statistically significant inverse association in three of the four studies, including the current one ( 12 , 16 ), while two of the three studies that examined lovastatin reported a statistically significant inverse association ( 13 , 16 ). Lovastatin was not marketed in the United Kingdom, so it could not be examined in the current study. It is possible that some variability in results could be because of genetic susceptibility in response to particular statins ( 36 ).

Confounding by contraindication related to liver disease has been raised as a possible explanation of inverse associations between statins and liver cancer ( 23 ). While hepatotoxicity is a rare complication ( 24 ), concern about its occurrence could result in biased prescribing patterns. Several studies have attempted to address this issue by conducting stratified analyses. In an analysis restricted to persons without liver disease, El-Serag et al. ( 17 ) found that statins remained statistically significantly inversely related to liver cancer. Similarly, after stratification on cirrhosis, Chiu et al. ( 12 ) found a dose-dependent reduction in risk among persons with and persons without cirrhosis. The current study found similar results in that there was a reduction in risk both among persons with liver disease and persons without liver disease. The association in persons with liver disease, in fact, appeared stronger. As recent reports suggest that concerns about prescribing statins to persons with liver disease may have been overstated ( 37 , 38 ), the current study suggests that persons at high risk of liver cancer also experience risk reductions with statin use. An analogous risk reduction was observed among persons with diabetes, lending further evidence in support of a cancer chemopreventive effect in a high-risk group.

Statins reduce serum cholesterol levels by inhibiting the rate-limiting enzyme in the production of cholesterol in the mevalonate pathway ( 39 ). By inhibiting 3-hydroxy-3-methylglutaryl coenzyme-A reductase, statins decrease hepatic cholesterol production, thereby increasing both low-density lipoprotein (LDL) receptor turnover and hepatic update of LDL from the circulation ( 40 ). The reduction in serum cholesterol may have a chemopreventive effect on liver cancer by lowering the cholesterol content of circulating lipid rafts that regulate a number of signaling pathways. Statins may also, via cholesterol reduction, regulate the function of a pathway important in carcinogenesis, the Hedgehog signaling pathway ( 41 ). It is also possible that statins exert anticarcinogenic effects via mechanisms unrelated to cholesterol level. In addition to cholesterol, the mevalonate pathway produces a number of products that are important for cellular function, including geranylgeranyl pyrophosphate (GGPP), farnesyl pyrophosphate (FPP), ubiquinol, dolichol, and Heme-A. Both GGPP and FPP activate proteins including Ras, and members of the Rab, Rac, and Rho families ( 42 ). By inhibiting production of GGPP and FPP, statins may also contribute toward the induction of apoptosis ( 43 ). Statins have also been shown to inhibit the activation of the proteasome pathway ( 44 ), thereby contributing to the maintenance of proteins that block cell cycling. Among persons infected with HCV, statins have also been demonstrated to inhibit viral replication, as GGPP is needed for HCV replication. In persons infected with HBV, a possible mechanism of action is simply the lowering of cholesterol levels, as cholesterol depletion impairs the ability of HBV to infect target cells ( 45 ).

The current study had several major strengths, including being one of the largest liver cancer studies to date in a western population. The study was conducted using a large, well-established, validated, longitudinal primary-care database that is known for accuracy of diagnoses, including cancer diagnoses, and completeness of pharmacy data. All information on diseases and drug exposures in the CPRD is recorded in the absence of a study hypothesis, thus the current study was not susceptible to recall bias. In addition, exposure was categorized by individual statins using medical record data that extended back an average of more than 12 years before the index date. All analyses were adjusted for a range of potential confounders, including BMI, smoking, alcohol abuse, HBV/HCV, diabetes, liver disease, and paracetamol use. By excluding case patients and control patients with less than two years of history in their medical record before the index date, the risk of incorrectly classifying statin use was minimized. In addition, the study was conducted in a country that has universal health care coverage, thus decreasing the chance that the results were biased because of failure to consider socioeconomic status. Statin usage has been reported to be greater among persons of higher socioeconomic status in some countries ( 46 ). Finally, two sensitivity analyses were conducted and each yielded consistent findings.

In contrast with its strengths, the current study also had several limitations. Although previous validation studies have reported that cancer diagnoses in the CPRD are accurate and complete ( 28 ), it is possible that some secondary liver cancers were erroneously included as verification of diagnoses via linkage to a cancer registry was not undertaken. To minimize this possibility, the current study excluded patients with prior cancer diagnoses (except nonmelanoma skin cancer) in at least the three years prior to the liver cancer diagnosis. In addition, the majority of the case patients (86.7%) had clinical codes consistent with a liver cancer diagnosis in their records. In a sensitivity analysis restricted to these case patients and their matched control patients, no material differences in results from the total population were found. In addition, it is likely that ascertainment of HBV and HCV status was not complete, as persons can be infected without being aware of it; race and ethnicity are not recorded uniformly in CPRD, so these variables could not be included as covariables. The majority (84%) of the UK population, however, is white, so any extrapolation to persons of other racial/ethnic groups should be done cautiously. Finally, the current study used existing records to examine the statins–liver cancer relationship, so it was only able to adjust for conditions as they were recorded in the database. A prospective study design would have permitted the more precise capture of medical conditions, but such a study would be prohibitively large and costly given that liver cancer is a rare outcome. For this reason, no prospective cohort study of statins and liver cancer has ever been mounted. The rarity of the outcome is also a hindrance to the conduct of a randomized controlled trial, as is the ubiquity of statin use in the adult population. To conduct a trial, it is likely that many persons in the non-intervention arm would have to be taken off a class of drugs that has been demonstrated to reduce serum cholesterol levels, thereby undoubtedly raising ethical concerns.

In conclusion, the results of the current study suggest that use of statins among persons at high risk of developing liver cancer, even in low-risk settings, may have a net cancer protective effect.

Funding

This research was supported by the Intramural Research Program of the National Institutes of Health.

The study funder had no role in design of the study, the collection, analysis, or interpretation of the data, the writing of the manuscript, nor the decision to submit the manuscript for publication.

References